That's all we have space for in a node record.

Also MDB_MAXKEYSIZE is redefinable at compile time.

Forgot this in 98fe3390

Very large single transactions will fail. It's not just a problem when
nested transactions are used. We could make this dynamically sized,
but I'm not sure what the point is.

If we're operating on the freelist, see if our current pghead
can satisfy the request before giving up and using new pages.

Was leaving it 0 initially.

Use the latest meta page

Allow both increasing and decreasing the environment size.
But don't allow decreasing below the currently occupied space.

These ops could fail if the map is full.

Don't try to reclaim overflow pages while operating on
the freelist (for now). The circular dependencies are much like
the single-page case, but worse. Maybe look into this in the
future, but it's not absolutely necessary now.

Keep list sorted if it grows during a write. Don't free pghead until
we're sure our writes are all finished.

Just write in 2^31 sized chunks

Avoid possible symbol clashes with MDB-Tools libmdb

mdb_cursor_sibling() no longer pops cursor before returning.

Avoid naming conflicts with other mdb* packages. Bump version to 0.9.5

With MDB_MAPASYNC, the API provided no way to ensure full sync.

It returned 0, which is the freelist DBI and is valid for a few uses.

This allows for other flag values in the future.

Let the user specify the number of databases he will actually
use. Adding 2 for FREE_DBI and MAIN_DBI is an internal matter.

mdb_env_open() with these flags did not open a synchronizing
file descriptor.  Thus disabling them later did not work.

Drop mmap()'s MAP_FIXED flag, so it returns another address instead
of unmapping existing overlapping pages. Verify the returned address.